The AI rollout framework

Most AI initiatives fail not on the technology but on the strategy behind it. A tool is bought, a few people try it out — and three months later nobody talks about it any more. We developed a framework from our AI adoption work. This chapter is the map for it.

The framework at a glance

A successful rollout stands on two foundations — and then moves in two directions. Leave one out and you stay stuck in pilot limbo.

• Foundation 1 — AI strategy: a clear rollout plan and champions as multipliers. Not "who gets the tool" but "how does the whole organisation learn".
• Foundation 2 — AI platform: a platform that becomes the operating system for work — chat, assistants, workflows, knowledge and integrations in one secure place, instead of ten island solutions.
• Go broad takes AI to the breadth — to every employee.
• Go deep takes AI to the depth — into every system and every process.

The five stages of your AI journey

Maturity does not arrive overnight. Almost every organisation goes through the same five stages — the art is going through them deliberately and in the right order.

The first two stages are go broad, the last three go deep. The use-case map shows what people really do with AI at stage 2 — and measuring impact works out what that is worth.

Stage 1 — the three-month pilot programme

A pilot is not a tool test but a guided process with a clear goal: real, measurable use cases in three months. Four building blocks carry it:

Go broad — reaching every employee

AI creates its value in breadth, not in a lighthouse project. "Go broad" has two pillars: enabling people — and anchoring the learning directly in the tool.

• People enablement: champions as multipliers, community & best practice, events & training, expert sparring. → Change management
• Platform adoption: learning in the tool — a prompting coach, curated top use cases and prompts, guided starts. → Enablement & training

Go deep — integrating every system

For AI to take on real work, it has to reach your systems. Two routes, deliberately separated by risk:

• RAG search index (read-only): AI searches your knowledge — documents, structured data, specifications — without changing anything.
• MCP integrations (read & write): AI acts in your systems, with clean permissions. → Automation and governance & security go into depth here.

What matters

Where to start

Start where the pain is large and the risk is small: recurring writing work, research, summaries. These early wins create the confidence for the more demanding stages. Look in the use-case map for where your team gains most — and take the matching assistants, workflows and prompts from the solution library.